Hello all,
I am investigating a hypothesis for cognition decline in older populations. Based on theory and literature, I think my model needs a four-way interaction term. The likelihood ratio test and AIC test results show the four-way interaction model is better than the main effect model, two-way interaction model, and three-way interaction model (AIC difference of 14 AIC unit, p=0.00001 for the LRT tests). The results from the four-way interaction model also meet my expectation.
But when I presented my study findings to my colleagues, some raised concerns about the power. My study leveraged a longitudinal sample of 900 individuals over six survey waves with a total of 4300 person-year observations. I used a linear mixed effect model. Some of my colleagues questioned the power of my research and said my sample size was insufficient for analyzing the four-way interaction model.
My question is, theory vs. power, which is more important? Both theory and past literature support the theory of the four-way interaction model. However, the sample is the only and the best one I can get for now to study my research question. Based on these conditions, should I give up doing this research and wait until a larger sample becomes available (i have to wait for a very long time)? Or it's worth to keep doing this topic and report results?
I am looking forward to hearing your comments.
I am investigating a hypothesis for cognition decline in older populations. Based on theory and literature, I think my model needs a four-way interaction term. The likelihood ratio test and AIC test results show the four-way interaction model is better than the main effect model, two-way interaction model, and three-way interaction model (AIC difference of 14 AIC unit, p=0.00001 for the LRT tests). The results from the four-way interaction model also meet my expectation.
But when I presented my study findings to my colleagues, some raised concerns about the power. My study leveraged a longitudinal sample of 900 individuals over six survey waves with a total of 4300 person-year observations. I used a linear mixed effect model. Some of my colleagues questioned the power of my research and said my sample size was insufficient for analyzing the four-way interaction model.
My question is, theory vs. power, which is more important? Both theory and past literature support the theory of the four-way interaction model. However, the sample is the only and the best one I can get for now to study my research question. Based on these conditions, should I give up doing this research and wait until a larger sample becomes available (i have to wait for a very long time)? Or it's worth to keep doing this topic and report results?
I am looking forward to hearing your comments.
